R

Ta

TT

a

AA

KCDEPT

C

0d

Journal of Chromatography A, 1217 (2010) 2703–2715

Contents lists available at ScienceDirect

Journal of Chromatography A

journa l homepage: www.e lsev ier .com/ locate /chroma

eview

he composition-explicit distillation curve technique: Relating chemical analysisnd physical properties of complex fluids�

homas J. Bruno ∗, Lisa S. Ott, Tara M. Lovestead, Marcia L. Huberhermophysical Properties Division, National Institute of Standards and Technology, Boulder, CO, USA

r t i c l e i n f o

rticle history:vailable online 17 November 2009

eywords:hemical analysisistillation curventhalpy of combustionetroleomicsrace analysis

a b s t r a c t

The analysis of complex fluids such as crude oils, fuels, vegetable oils and mixed waste streams posessignificant challenges arising primarily from the multiplicity of components, the different properties ofthe components (polarity, polarizability, etc.) and matrix properties. We have recently introduced ananalytical strategy that simplifies many of these analyses, and provides the added potential of linkingcompositional information with physical property information. This aspect can be used to facilitate equa-tion of state development for the complex fluids. In addition to chemical characterization, the approachprovides the ability to calculate thermodynamic properties for such complex heterogeneous streams.The technique is based on the advanced distillation curve (ADC) metrology, which separates a complexfluid by distillation into fractions that are sampled, and for which thermodynamically consistent tem-peratures are measured at atmospheric pressure. The collected sample fractions can be analyzed by anymethod that is appropriate. The analytical methods we have applied include gas chromatography (withflame ionization, mass spectrometric and sulfur chemiluminescence detection), thin layer chromatogra-

phy, FTIR, corrosivity analysis, neutron activation analysis and cold neutron prompt gamma activationanalysis. By far, the most widely used analytical technique we have used with the ADC is gas chromatog-raphy. This has enabled us to study finished fuels (gasoline, diesel fuels, aviation fuels, rocket propellants),crude oils (including a crude oil made from swine manure) and waste oils streams (used automotive andtransformer oils). In this special issue of the Journal of Chromatography, specifically dedicated to extrac-tion technologies, we describe the essential features of the advanced distillation curve metrology as an analytical strategy for complex fluids.

Published by Elsevier B.V.

ontents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27041.1. Advanced distillation curve method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2704

2. Applications of the ADC method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27062.1. Volatility and detailed chemical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27062.2. Hydrocarbon type analysis—aviation fuels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27072.3. Volatility and energy content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27082.4. Tracking selected components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27092.5. Detection of azeotropes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27102.6. Study of azeotropes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27112.7. Volatility and chemical stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2711

2.8. Volatility and corrosivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3. Thermodynamic modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

� Contribution of the United States government; not subject to copyright in the United∗ Corresponding author. Tel.: +1 303 497 5158; fax: +1 303 497 5927.

E-mail address: bruno@boulder.nist.gov (T.J. Bruno).

021-9673/$ – see front matter. Published by Elsevier B.V.oi:10.1016/j.chroma.2009.11.030

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2712

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2712

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2715. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2715

States.

2 togr.

1

opsTsa[fapicfim

opmTmypampota[

avtitfvtotblaDmahtlMsc

coatabsdmd

704 T.J. Bruno et al. / J. Chroma

. Introduction

This special issue of the Journal of Chromatography provides thepportunity to highlight the importance of separation/extractionrocesses in general, and how the specific techniques of chemicaleparations can be used to solve laboratory and industrial problems.he importance of separation processes cannot be understated,ince nearly 40% of the cost of any chemical product is directlyttributable to the separation processes used in their production1]. These processes include distillation, extraction, adsorption, dif-usion and zone refining, to mention just a few. For industrialpplications, distillation has long been the dominant separationrocess, in terms of the overall number of units implemented, and

n terms of total capital investment. The relative simplicity and effi-iency, and the long term experience base makes distillation therst choice in the chemical processing industry, even when otherethods might be viable.Aside from the chemical production aspects, the dependence

f separations techniques on intermolecular interactions has alsorovided the metrology to study chemical properties, and to deter-ine how such properties relate to the constituents of mixtures.

he application of chromatographic methods for physicochemicaleasurements, for example, has been well known for nearly 50

ears [2]. Thermodynamic parameters measured by chromatogra-hy have been validated by other techniques such as spectroscopynd calorimetry. In a similar manner, we can use distillation as aeasurement method, one that is especially applicable for com-

lex fluids. Here, the relationship between composition on thene hand and vapor liquid equilibrium on the other (both con-rolling factors in distillation) furnishes us with a bridge betweennalytical chemistry and thermophysical property measurement3,4].

Since complex, multi-component fluid mixtures vaporize overrange of temperatures, the only practical avenue to assess the

apor liquid equilibrium (VLE) is the measurement of the distilla-ion or boiling curve [5]. The classical distillation curve of a fluids a graphical depiction of the boiling temperature of the mix-ure plotted against the volume fraction distilled. This volumeraction is usually expressed as a cumulative percent of the totalolume. One most often thinks of distillation curves in the con-ext of petrochemicals and petroleum refining, but such curves aref great value in assessing the properties of any complex mix-ure [6–9]. Indeed, the measurement of distillation curves haseen part of complex fluid specifications for a century (typically

isted in specifications and data sheets as the fluid volatility),nd they are inherent in the design and application of all fuels.espite this importance, the standard methods for the measure-ent of such curves have been plagued with systematic uncertainty

nd bias, and an absence of any link to fluid theory [10]. Thisas lead many petroleum scientists and engineers to considerhe measurement of distillation curves to be virtually meaning-ess, valuable only because everybody has done it the same way.

oreover, the standard metrology has always ignored the compo-itional aspects that are so important; distillation is really all aboutomposition.

We recently introduced an improved method, called theomposition-explicit or advanced distillation curve (ADC) metrol-gy, as a means to characterize complex fluids [11–14]. The ADCpproach addresses many of the shortcomings of the classical dis-illation methods described above. Most important, we incorporatecomposition-explicit data channel for each distillate fraction (for

oth qualitative, quantitative and trace analysis). Sampling verymall distillate volumes (5–25 �L) yields a composition-explicitata channel with nearly instantaneous composition measure-ents. Chemical analysis of the distillate fractions allows for

etermination of how the composition of the fluid varies with vol-

A 1217 (2010) 2703–2715

ume fraction and distillation temperature, even for complex fluids.These data can be used to approximate vapor liquid equilibrium ofcomplex mixtures, and present a more complete picture of the fluidunder study. The ADC approach provides consistency with a cen-tury of historical data, an assessment of the energy content of eachdistillate fraction, and where needed, a corrosivity assessment ofeach distillate fraction. Suitable analytical techniques include gaschromatography with either flame ionization detection (GC-FID) ormass spectral detection (GC-MS), element specific detection (suchas gas chromatography with sulfur or nitrogen chemiluminescencedetection, GC-SCD or GC-NCD), Karl Fisher coulombic titrimetry,refractometry, and Fourier transform infrared spectrometry (FTIR)[15,16].

Another advantage of the ADC approach is that it provides tem-perature, volume and pressure measurements of low uncertainty,and the temperatures that are obtained are true thermodynamicstate points that can be modeled with an equation of state. In fact,we have used the ADC method to develop chemically authenticsurrogate mixture models for the thermophysical properties of acoal-derived liquid fuel, a synthetic aviation fuel, S-8, and rocketpropellants RP-1 and RP-2 [11–14,17–19]. The ability to couplethe compositional data with the thermal data gives us access tonumerous material dependent quantities, and the ability to relatethem to the mixture volatility. We will illustrate this in the selectedexamples that follow.

1.1. Advanced distillation curve method

The apparatus and procedure for the measurement of the com-position ADC have been discussed in detail elsewhere; only a briefdescription will be provided here [11,13]. The apparatus is depictedschematically in Fig. 1. The stirred distillation flask is placed inan aluminum heating jacket contoured to fit the flask. The alu-minum jacket serves to integrate out temperature gradients inall but the vertical direction, in which direction a gradient mustexist to provide mass transfer. The jacket is resistively heated,controlled by a model predictive PID controller that applies aprecise thermal profile to the fluid [14]. Three observation portsare provided in the insulation to allow penetration with a flex-ible, illuminated borescope. The ports are placed to observe thefluid in the boiling flask, the vapor space at the top of the boilingflask, and the vapor in the distillation head (at the bottom of thetake-off).

Above the distillation flask, a centering adapter provides accessfor two thermally tempered, calibrated thermocouples that enterthe distillation head. Calibration is normally done with a triplepoint cell that can be traced to a NIST standard. One thermocou-ple (T1) is submerged in the fluid and the other (T2) is centeredin the head at the low point of distillate take-off. Also in the headis a stainless steel capillary tube that provides an inert gas blan-ket for use with thermally unstable fluids. Distillate is taken off theflask with the distillation head, into a forced-air condenser that ischilled with a vortex tube [20–22]. The vortex tube is a device thatproduces a stream of hot and cold air from an ordinary shop airsource. We use air as the cooling medium because it is easily con-trolled, and it avoids the problems associated with a water flow.Water cooled condensers can cause glassware cracking due to thehigh heat capacity of water, and always involve the potential forleaks.

Following the condenser, the distillate enters a new transferadapter that allows instantaneous sampling of distillate for anal-

ysis. The flow path of the distillate is focused to drop into a 0.05 mL“hammock” that is positioned directly below the flow path. A crimpcap or screw cap fixture is incorporated as a side arm of the adapter.This allows a replaceable silicone or Teflon septum (of the type usedfor chromatographic automatic sampler vials) to be positioned in

T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715 2705

F stillatis

lomtt

lsgssdenpw

fef(

ig. 1. Schematic diagram of the overall apparatus used for the measurement of dihown in the lower half of the figure.

ine with the hammock. The distance from the crimp cap to the basef the hammock is suited to the needle length of typical gas chro-atographic syringes. To sample the distillate, one simply positions

he chromatographic syringe, preferably equipped with a bluntipped needle, in the well of the hammock.

When the sample leaves the adapter, it flows into the calibrated,evel-stabilized receiver for a precise volume measurement. Con-tructed of glass, this receiver consists of a central volume thatradually decreases in diameter at the base, and connects to amall-diameter side arm sight glass that is calibrated. The side armtabilizes the fluid level for a precise volume measurement as theistillation proceeds. The large inner volume and the sight glass arenclosed in a water jacket that contains a thermometer and a mag-etic stir bar for circulation. When surface tension effects becomeroblematic (as with mixtures of polar and nonpolar constituents),e use a receiver with volumes of equal diameters.

Since the measurements of the distillation curves are per-ormed at ambient atmospheric pressure (measured with anlectronic barometer), temperature readings are usually correctedor what should be obtained at standard atmospheric pressure1 atm = 101.325 kPa) [23]. This adjustment is done with the NIST-

on curves. Expanded views of the sampling adapter and the stabilized receiver are

modified Sydney Young equation [23–27]. The typical temperatureuncertainty is less than 0.3 ◦C (with 2�), the volume uncertaintyis 0.05 mL, and the uncertainty in the pressure measurement is0.003 kPa.

To measure a distillation curve, fluid (40–200 mL) is placed inthe distillation flask and the heating profile begins. We alludedto this earlier when mention was made of the model predictivetemperature controller [14]. The thermal profile typically has thesigmoidal shape of a distillation curve, but continuously leads thefluid by ≈20 ◦C. Thus, the typical vaporization behavior of the fluidis approximated in a model, and applied to the surroundings of theflask. This is done to ensure a constant mass flow rate through theapparatus.

For each ADC measurement, we can record a data grid consistingof: Tk, the temperature of the fluid (measured with T1), Th, the tem-perature in the head (measured with T2), the corresponding fluid

volume, the elapsed time, and the external (atmospheric) pressure.Along with these data, one withdraws a sample for detailed anal-ysis. This procedure provides access to the detailed composition,energy content, corrosivity, etc., corresponding to each datum inthe grid.

2706 T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715

F y planr ass sps

2

2

iflj

sTcrlattfpmaetfl

tstlGaqlssa

ig. 2. A distillation curve for RP-1 showing Tk against volume fraction in the x–epresented as retention time against peak intensity. Inset “a” above shows the mulfur chromatographic peak.

. Applications of the ADC method

.1. Volatility and detailed chemical analysis

A detailed chemical analysis coupled with physical propertynformation is very helpful if not essential in QA/QC for complexuids. Examples can be drawn from the kerosenes that are used for

et fuels and rocket propellants.While modern rocket motors can operate on either a liquid or a

olid fuel package, the former is more easily controlled and flexible.his led to the development of RP-1 kerosene in the 1950s, whichontinues to be widely used [28]. The desire in recent years to useocket motors many times has led to reformulations of RP-1 withow sulfur, olefin and aromatic content. Reformulation has requiredreassessment of the physical properties, for which we have used

he ADC metrology. We show in Fig. 2 a distillation curve of RP-1hat has the composition measurement superimposed [12]. First,ocusing on the plot of Tk against volume fraction, we note that thelot shape is a subtle sigmoid, characteristic of a complex fluid withany components. ADC data such as these are used in the design

nd specification of many engine operational parameters, and inquation of state-based model development. Since the Tk data arehermodynamic state points, the plot represents a cut through theuid phase diagram that has theoretical meaning.

The composition-explicit channel provides additional informa-ion for the data grid. In Fig. 2, the composition is shown forelected temperature–volume pairs as measured by GC-MS. Addi-ional detail is shown in the inset, where the mass spectrum of theargest peak is identified as n-dodecane. Clearly, the application ofC-MS can be used to any degree of detail that is desired. We canlso use element specific detection in many cases to answer specific

uestions. As discussed above, RP-1 has been reformulated to have

ower sulfur content, in order to decrease corrosivity and metal ero-ion in the engines. Thus, the application of sulfur analysis with aulfur chemiluminescence detector is of critical importance. Thisnalysis is also shown in the inset, allowing a breakdown of the

e, and the composition as measured by gas chromatography along the “z” axis,ectrum of the major peak of the 40% fraction, n-dodecane; inset “b” shows a total

sulfur budget on the basis of fraction volatility. This compositionalinformation is now joined with a temperature grid measurementdiscussed above; the temperature, pressure and composition canall be modeled with an equation of state, as discussed later.

We can provide a more detailed picture of the results from thecomposition-explicit channel in the examination of the distillatefractions of a sample of Jet-A (flash point ≈38 ◦C, freezing tem-perature −40 ◦C), the major gas turbine aviation kerosene usedcommercially in the United States, with a consumption of 800 bil-lion liters in 2006. To ensure an adequate supply for commercialmarkets, the overall specifications of Jet-A are relatively wide interms of thermophysical properties. This is reflected in sometimeswidely varying distillation curves that can be measured for accept-able, in-specification fluids. It can therefore be difficult to definea “typical” sample of Jet-A, however in earlier work we measuredthe distillation curves of a composite sample of Jet-A. This samplewas prepared by combining aliquots from approximately ten sep-arate lots of Jet-A. We show in Fig. 3 a series of chromatogramsmeasured with a FID of the distillate fractions for this compos-ite sample of Jet-A [29,30]. The time axis is from 0 to 12 min foreach chromatogram, and the abundance axis is presented in arbi-trary units of area counts (voltage slices). It is clear that althoughthere are many peaks on each chromatogram (30–40 major peaksand 60–80 minor and trace peaks), these chromatograms are muchsimpler than those of the neat fluids, which can contain 500–800major peaks. At the very start of each chromatogram is the solventfront, which does not interfere with the sample. One can follow theprogression of the chromatograms in Fig. 3 as the distillate fractionbecomes richer in the heavier components. This figure illustratesjust one chemical analysis strategy that can be applied to the dis-tillate fractions. It is possible to use any analytical technique that

is applicable to solvent borne liquid samples that might be desir-able for a given application. With this approach, we can track theappearance of the lightest components early in the distillation,and how these components disappear as the heavier componentsgrow in.

T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715 2707

F arbitd

2

etecoc(cciog

mfli

ig. 3. Chromatograms of distillate fractions of a typical Jet-A sample presented inetails of the chromatography are discussed in the text.

.2. Hydrocarbon type analysis—aviation fuels

It is not always necessary to include a detailed analysis forach line of the data grid. It is often of value to simply classifyhe chemical families that are present. This is possible with a vari-ty of methods. We routinely apply a mass spectrometric moietylassification method (similar to ASTM Method D-2789), in whichne characterizes hydrocarbons into six types: paraffins, monocy-loparaffins, dicycloparaffins, alkylbenzenes, indanes and tetralinsgrouped), and naphthalenes. While not without its limitations, andertainly not the only such test used for gross characterization, itan be used reliably as a comparative tool for evaluations betweenndividual batches of complex fluids. We illustrate the applicationf this approach to the comparison of the distillation curve data

rid of two aviation turbine fuels, JP-8 and S-8 [31].

JP-8 is the major turbine fuel currently used by the United Statesilitary (MIL-DTL-83133), a kerosene fraction that has a higher

ash point than the main military predecessor, JP-4. JP-8 was firstntroduced at NATO bases in 1978 and is currently the US Air Force’s

rary units of intensity (from a flame ionization detector) plotted against time. The

primary fuel, and the primary fuel for US Navy shore-based avia-tion. JP-8 is very similar to Jet-A-1, the most common commercialgas turbine fuel in Europe, with the major differences being in theadditive package. Note that Jet-A and Jet-A-1 differ in an additivethat decreases the freezing point of Jet-A-1 to −47 ◦C. JP-8 alsotypically contains an icing inhibitor, corrosion inhibitor/lubricityenhancer and anti-static additive. Significant variability is observedin the composition of JP-8 (and the Jet-A fluid discussed above).Much of this variability is intentional, to allow for an ample supply.Additional variability found in JP-8 results from the additive pack-age, which is often splash blended at the flight line. Understandingthe composition and variability has always been important, andis currently becoming even more critical, as detailed in the nextparagraph.

There is a desire in the United States defense community to uti-lize JP-8 as the main battlefield fuel for all vehicles, not only foraviation applications but also for ground based forces. Environ-mental concerns and the potential of disruptions in supply haveled to the development of synthetic aviation fuels. Synthetics are

2708 T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715

nt in a

o8p4tht5

mciadfigdWfottatenab

2

ts

Fig. 4. Plots showing the distribution of chemical families prese

f interest as extenders and even as drop-in replacements for JP-. One such fuel made from natural gas (with the Fischer Tropschrocess) is designated as S-8 (the “S” referring to synthetic; CAS No.37986-20-4) [12]. This fluid is a hydrocarbon mixture rich in C7o C18 linear and branched alkanes, with few if any aromatics. Itas a flash point range of between 37.8 and 51.8 ◦C, an autoignitionemperature of 210 ◦C, and explosive limits in air between 0.7 and(v/v).

A key engineering parameter to facilitate application of S-8, andixtures of S-8 with JP-8 is the ADC. By adding the chemical family

lassifications to the data grid, the distillation curve becomes morenformation-rich. In Fig. 4, we present the classification results, asfunction of distillate cut, for typical batches of JP-8 and S-8. Theifferences are striking. We note that S-8 has a high level of paraf-nic species and very few aromatics (as expected from its naturalas feedstock), and JP-8 has a high aromatic content decreasing asistillation proceeds (as expected from its petroleum feedstock).e have found that this behavior is typical of kerosenes and diesel

uels [32–36], and is in contrast to that of gasolines, where onebserves the aromatics to increase and aliphatics to decrease ashe distillation proceeds [37]. The importance of this characteriza-ion technique stems from the required overall specifications forviation fuels. Aromatics must be present to some extent in avia-ion fuels in order to meet the density specification. If present inxcess, however, aromatics will produce soot and an overly lumi-ous flame. The ability of the ADC to combine a check on theromatic content and the physical property information providedy the boiling range is thus efficient and cost-effective.

.3. Volatility and energy content

The ability to apply a detailed quantitative analysis to each dis-illate fraction offers the potential of assessing thermal propertiesuch as energy content of a fuel. If the enthalpy of combustion

viation fuels JP-8 and S-8, as determined by mass spectrometry.

is known (or predictable) for the components of a mixture, thecomposite enthalpy of combustion of a mixture of these compo-nents can be derived (neglecting the enthalpy of mixing). We havedemonstrated how this can be applied to the distillate fractionscorresponding to the data grid of the distillation curve [38]. It isnot necessary to identify all the components of the fraction; asubstantial subset of the major constituents is adequate, and theuncertainty caused by the use of a subset is negligible (that is, lessthan the uncertainty resulting from the pure component enthalpy).

Continuing with the comparison of aviation fuel properties, onecan appreciate that a major interest is the energy content. More-over, since droplet combustion occurs inside the combustors ofmodern turbofan engines, the energy content as a function of dis-tillate cut is very important. The shrinkage of a droplet in thecombustor is accompanied by compositional changes, mirrored bythe distillation curve. We illustrate the application of the ADC mea-surement of energy content with a comparison of different samplesof Jet-A [29]. The ADC was applied to three different batches of Jet-A(designated numerically as 3638, 3602 and 4658) that are thoughtto represent the composition gamut very well. The sample labeled4658 is the composite mentioned earlier. It is therefore consideredto be the most representative of the three samples. The samplelabeled 3638 was known to be unusual in that the aromatic con-tent was lower than typical batches, while that labeled 3602 wasunremarkable and typical. We noted a divergence in the distilla-tion curves of these three fluids at the 70% fraction, so a quantitativeanalysis was done at this fraction for each fluid. We then applied ourmethod to determine the enthalpy of combustion of this fraction,the results of which are shown as a histogram in Fig. 5, along with

a comparison to the synthetic fluid made from natural gas, S-8. Wewere surprised to note a significant spread in the enthalpy valuesamong these fluids. The mixed sample shows the highest energycontent, while the atypical fluid 3638 shows the lowest. The com-bination of the distillation data grid with the composition analysis

T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715 2709

Fte

ao

2

ppak(ocaW

sfsimtatcD

F((

ig. 5. The composite enthalpy of combustion of the 70% distillate fraction forhree samples of Jet-A and the sample of S-8. The fluids are presented in increasingnthalpy of combustion.

nd the enthalpic analysis permits a more complete understandingf the fuel properties, and how they relate to composition.

.4. Tracking selected components

Finished fuels often incorporate additives for specific pur-oses, including oxygenating agents, antiknock agents, extenders,reservatives, antifoam and lubricity agents, and detergents. Somedditives were mentioned above in our discussion of aviationerosenes. While some of these are present at trace levels, othersespecially oxygenates and extenders) are added in concentrationsf 10% or higher. The development of models for the thermophysi-al properties of such fluids requires explicit knowledge of how thedditives change the fundamental properties such as the volatility.e can use the ADC to unify these two important parameters.Oxygenates added to gasoline to reduce carbon monoxide emis-

ions are familiar, but various oxygenates have been added to dieseluel to decrease (or eliminate) particulate formation. We have mea-ured numerous gasoline and diesel fuel mixtures with oxygenates,ncluding synthetics and biomass derived fluids [32,33,35]. Since

any engine operation and environmental parameters depend on

he distillation curve, the ability to relate the changing compositionnd actually model the fluid behavior is critical. In Fig. 6, we presenthe ADC results for mixtures of diesel fuel with three different con-entrations of diethyl carbonate (DEC), a promising oxygenate [32].EC is used extensively as an ethylating agent in organic synthe-

ig. 6. The distillation curves of diesel fuel and diesel fuel with 10, 20, and 30%v/v) of diethyl carbonate (DEC). The inset shows the concentration profile (in %mass/mass)) of the additive as a function of distillate cut.

Fig. 7. Distillation curve of avgas 100LL with the enthalpy of combustion in inset(a), and the concentration of tetraethyl lead (TEL) in inset (b), both as a function ofdistillate cut.

sis (for example, it is used in the synthesis of the anticonvulsantdrug Phenobarbital), and it is also used extensively as a solvent inthe textile industry. It is biodegradable and insoluble in water. Wenote that in mixtures with diesel fuel, the DEC causes a significantinflection to lower temperatures. It is more volatile than most ofthe diesel fuel components, and even at a starting concentrationof 30% (v/v), it has been removed from the diesel fuel by the 0.5distillate fraction, at which point the distillation curves of the DECmixture is approaching (but not merging with) that of diesel fuel.This is reflected in the inset, in which a quantitative analysis of theDEC in the distillate is provided, having been measured by GC-FID.Especially interesting is that the additive, although removed by adistillate fraction of 0.5, still apparently has an effect on the VLElate in the distillation. The vaporization of the lighter components ofdiesel fuel, which would ordinarily be found early in the distillationcurve, is delayed by the presence of the additive. We will discussthis in more detail later in the section on thermodynamic modeling,however it is precisely this combination of temperature and com-position information that permits a more complete understandingof the behavior of such complex mixtures.

Another instructive example of how we can track an additivethrough the distillation curve (but this time an additive concen-tration approaching the trace concentration level) comes from themeasurement of the commercial aviation gasoline, avgas 100LL.Although motor fuels used today in the United States and Europe donot contain lead additives, much general aviation gasoline (avgas100LL) still contains tetraethyl lead (TEL, CAS No. 78-00-2). SinceTEL was banned from motor gasoline, avgas is now one of thelargest contributors of lead in the atmosphere in many locations.Significant efforts have been made to develop a low cost, lead freealternative fuel to replace avgas 100LL for aircraft that use pistonengines. The examination of avgas 100LL with the ADC provides theopportunity to ultimately develop an equation of state for avgas,and to track the presence of the lead compound through the fullrange of the distillation curve [39]. In Fig. 7, we apply the ADCto avgas 100LL [39]. The y-axis presents the thermodynamicallyconsistent temperatures. In inset (a) we present the enthalpy ofcombustion as a function of distillate cut (from a quantitative anal-ysis of each fraction). This allows the energy content to be relatedto the other fuel properties as a function of distillate cut. In inset

(b), we present the composition profile of TEL as a function of dis-tillate cut, which comes from specific trace analysis applied to thedistillate cuts. We note that there is far more TEL in later distillatefractions.

2710 T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715

F f methm

2

srt

acmcaabtttipoa[kTmami4m

F2

ig. 8. Distillation curves of (a) 91 AI gasoline and (b) 91 AI gasoline with 15% (v/v) oethanol and the some components of gasoline.

.5. Detection of azeotropes

Azeotropic mixtures are among the most fascinating and at theame time the most complicated manifestations of phase equilib-ium. They also play a critical role in many industrial processes (andhe resulting products), especially separations.

As we noted earlier, the ADC measures two temperatures, Tknd Th. Typically, during the measurement of a complex, multi-omponent fluid, the Tk measurement is higher than the Theasurement by several (5–15) degrees Celsius. This must be the

ase, since the mass transfer driving force comes from the temper-ture differential between the kettle and the head. If one performsn ADC measurement on a pure fluid, the temperature differenceetween Tk and Th is very small, no more than 0.1 ◦C (often less);he composition is not changing during the distillation. Moreover,he curve for a pure fluid is flat with zero slope. We would expecthis difference in temperature differential and slope to be reflectedn the distillation of an azeotrope, since where azeotropic pairs areresent, the mixture behaves as a pure fluid. Mixtures of gasolinexygenates in fact show this behavior, since the lower alcohols formzeotropes with many of the hydrocarbon components in gasoline37]. In Fig. 8a, we show the distillation curves of a 91 AI (anti-nock index) premium, winter grade gasoline, presented in Tk andh. This fuel has no added oxygenate. We note for this complex,ulti-component fluid that Tk is always higher than Th by an aver-

◦

ge of 6.2 C. In Fig. 8b we show the same gasoline with 15% (v/v)ethanol. Two features are noteworthy. First we observe a flatten-

ng of the curve for distillate volume fractions up to approximately0%, relative to that for the straight gasoline. This persists until theethanol has been distilled out of the mixture. Second, we also note

ig. 9. (a) The distillation curves of gasoline in mixtures with 1-butanol at concentration0% mixture, plotted in Tk and Th , showing the azeotropic convergence.

anol. The azeotropic convergence is caused by pairs of azeotropes forming between

the convergence of Tk and Th in this region, which we have calledthe azeotropic convergence. Here, the difference between Tk and Thaverages 0.3 ◦C, while subsequent to the azeotropic inflection, thedifference increases to an average of 8.6 ◦C.

Sometimes the azeotropic convergence is not as dramatic as inthe case of gasoline + methanol. This was illustrated in our measure-ments on mixtures of gasoline with the butanols [40]. This workstemmed from recognition of the many disadvantages of ethanolin fuel blends (including corrosivity toward ferrous metals, swellingof common elastomers used as seals in fuel systems, degradationof transfer lines, water absorption, phase separation, and a signif-icantly lower energy content than typical gasoline) [32]. Mixtureswith the butanols have been suggested as a possible alternative toavoid some of these difficulties. We applied the ADC to the fourbutanols, at mixture concentrations of 10, 20 and 30% (v/v). As anexample, we show the curves for 1-butanol in Fig. 9, and with themwe also show the Tk–Th behavior for the 20% (v/v) mixture. We notethat a convergence appears in the middle of the distillate range,corresponding with the vaporization of the 1-butanol. This resultsfrom the azeotropic binaries that occur between 1-butanol and thecomponents of gasoline.

It might at first be surprising that the azeotropic convergenceis very subtle, since our observation of the convergence with gaso-line + methanol was very dramatic. In that case, in the azeotropicregion, Tk and Th converged to the extent that they overlaid on one

another. The more subtle behavior here can be understood, how-ever, by the respective phase diagrams of the butanols with theconstituents of gasoline. The temperature of the azeotropic statepoint produced by the addition of methanol to a hydrocarbon is typ-ically rather far from the boiling temperatures of the hydrocarbons

s of 10, 20, and 30% (v/v). (b) The distillation curve of the gasoline + 1-butanol for a

T.J. Bruno et al. / J. Chromatogr. A

Faft

ta4toil

2

mtbib(idlTimpT6igpfl

t0cavca(acxd

channel of the ADC. We note that the stabilization effect is great-

ig. 10. A plot of the distillation curve data for binary mixtures of ethanol + benzenet benzene mole fractions xb = 0.20, 0.40, 0.55, 0.70 and 0.80. The inset shows theamiliar T–x diagram for this mixture. The azeotrope exists at a benzene mole frac-ion of 0.55.

hemselves. For example, for cyclohexane, n-heptane, n-octane,nd toluene, the temperature displacements are 53, 39, 63 and7 ◦C, respectively. For 1-butanol with these same hydrocarbons,he displacements are 3, 7, 16 and 5 ◦C, respectively. Thus, the effectf 1-butanol on the mixture boiling points is clearly more subtle,llustrating the link between the ADC measurement and the vaporiquid equilibrium of the mixture.

.6. Study of azeotropes

We can take the examination of azeotropes to a more funda-ental level by examining some well known binary mixtures with

he ADC. One of the most well studied mixtures is the minimumoiling binary azeotrope is that formed by benzene and ethanol. It

s often presented in introductory texts as an instructional exampleecause of the striking features and structure of the phase diagramthe temperature differences are significant, the two-phase regions large, and the azeotrope occurs nearly at the mid-point of the T–xiagram). This mixture is also industrially important in the formu-

ation and design of oxygenated and reformulated gasolines. The–x phase diagram of this binary, shown in the inset of Fig. 10,s anchored on the left side by the pure ethanol point (at a nor-

al boiling temperature of 78.4 ◦C), and on the right side by theure benzene point (at a normal boiling temperature of 80.1 ◦C).he bubble and dew point curves meet at the minimum located at8.2 ◦C. Centered about the minimum on the bubble point curve

s a relatively flat region where the slopes in either direction areentle. These slopes become increasingly more pronounced as oneroceeds away from the azeotrope. The dew point curves proceedrom the azeotropic point to the pure component points in a moreinear fashion with relatively constant slope.

Distillation curves are presented in this figure for binary mix-ures with starting compositions of 0.20, 0.40, 0.55, 0.70 and.80 mol fraction of benzene (xb) [41]. We note that the distillationurves for the starting compositions xb = 0.20 and 0.40 converget a temperature of 78.9 ◦C, while those at xb = 0.70 and 0.80 con-erge at a temperature of 80.9 ◦C. These two different families ofurves, which begin with starting compositions on either side of thezeotrope, converge to the appropriate pure component; 78.9 ◦Cfor x = 0.20 and 0.40, converging to ethanol) and 80.9 (for x = 0.70

b bnd 0.80 converging to benzene). We note that the shapes of theurves for xb = 0.20 and 0.80 are initially far steeper than those forb = 0.40 and 0.70. This can be explained with reference to the T–xiagram. We note that the initial steepness of slope corresponds

1217 (2010) 2703–2715 2711

with the pronounced increase in slope of the bubble point curve.Where the T–x diagram is steep, the distillation curve is correspond-ingly steep. This shape gives an indication of the deviations fromRaoult’s law, with steeper curves indicating larger deviations. Forthe mixture starting at a benzene mole fraction of 0.55, we notethat the distillation curve is flat, behaving as a pure fluid at theazeotrope. We also note that the liquid and vapor compositions arethe same. The ADC thus provides a simple and rapid avenue to thestudy of azeotropic mixtures. Each such curve can be completedin an hour with relatively simple instrumentation, whereas the T–xdiagram would require many hours to measure in a specialized VLEapparatus.

2.7. Volatility and chemical stability

Biodiesel fuel has been the focus of a great deal of mediaattention and scientific research in the last several years as apotential replacement or extender for petroleum-derived dieselfuel. The major constituents (fatty acid methyl esters, FAMEs) ofpure biodiesel are generally relatively few, consisting mainly ofmethyl palmitate, methyl stearate, methyl oleate, methyl linoleate,and methyl linolenate. As a fuel for compression ignition engines,biodiesel fuel has several advantages (renewable, increased lubric-ity, non-carcinogenic, non-mutagenic, biodegradable, decreasedcarbon monoxide, unburned hydrocarbon, and particulate matteremission). There are also some serious disadvantages to biodieselfuel (increased NOx emissions, moisture absorption during storage,and chemical instability). The last item is especially problematicat higher temperatures, although the instability in storage hasreceived more attention. We found in earlier work on biodieselfuel (B100) that the thermal and oxidative instability of this fluidprevented the measurement of a distillation curve with our usualADC approach; discrepancies in temperature of up to 20 ◦C wereobserved between successive measurements. The addition of anargon gas sparge incorporated into the distillation flask eliminatedthe problem, and allowed the measurement of highly reproducibledistillation curves [42,43]. Since it is possible to quantitativelyassess the “tightening” of replicate distillation curve measurementupon the addition of the sparge, we can use this change as a meansof assessing the thermal and oxidative stability of the fluids beingmeasured. We used three statistical descriptors of the improvedcurve-to-curve repeatability: the average range in temperature,the average standard deviation in temperature, and the area sub-tended. We found that these measures correlated quantitativelywith improved thermal and oxidative stability and thus provide ameasure of stability.

We then used the ADC as described above to test the efficacyof stabilizing additives on sensitive fluids such as B100 [44]. Inparticular, we tested three hydrogen donor additives: tetrahydro-quinoline (THQ), t-decalin and tetralin (the classical donor solventis composed of a saturated ring attached to an aromatic ring).Hydrogen donors are fluids or solvents that are capable of pro-viding hydrogen to enable the conversion of heavier residuals intodistillable fractions. They act to cap aliphatic radicals formed attemperatures in excess of 300 ◦C, and typically form C2 and C3 alkylaromatic compounds. In Fig. 11, we present the distillation curve ofB100 stabilized with 1% (v/v) THQ, and note that the repeatabilityof three successive curves is approximately 1.6 ◦C. This is a sig-nificant improvement from the unstabilized fluid (although not assignificant as with the argon sparge). We track the concentration ofTHQ in the distillate in the inset, with the composition-explicit data

est earliest in the curve, when the concentration of THQ is highest.The THQ decomposes and also distills out of the mixture duringthe course of the distillation, and its effect naturally decreases.Of the three stabilizers examined, the THQ and t-decalin perform

2712 T.J. Bruno et al. / J. Chromatogr.

F(b

stp

2

Mtmo“oabsDcatadaot

iac[dovl3t

ig. 11. Three successive distillation curves of B100 biodiesel fuel with 1% THQstructure provided) stabilizing additive. The inset tracks the concentration (in ppmy mass) of THQ present in the distillate.

imilarly; tetralin also resulted in stabilization, but with less effec-iveness. Interestingly, this result parallels kinetics measurementserformed on aviation fuels stabilized with these additives [45–47].

.8. Volatility and corrosivity

Crude oil is an economic driving force in the developed world.any properties of crude oil (color, viscosity, and amount and

ype of impurities) are dependent on its source. Impurities of pri-ary concern are sulfur species, which are often corrosive. Crude

ils containing relatively few sulfur impurities are referred to assweet”; they are considered “sour” if they contain large amountsf sulfur impurities. The corrosivity of crude oil streams is alwaysn important issue, one that can account for serious financial lia-ilities by producers and refiners. The corrosivity of certain sulfurpecies in fluids is determined by ASTM test methods D-1838 or-130, the copper strip corrosion test (CSCT) [48–54]. A strip ofleaned, polished copper is placed in a vessel and then filled withn appropriate quantity of the fluid to be tested. The filled vessel ishen maintained at an elevated temperature for a predeterminedmount of time, and the strip is removed from the fluid and imme-iately rated by comparison with a lithographed standard. Therere four levels of increasing corrosion on the standard, with levelne corresponding to slight tarnishing and level four correspondingo severe corrosion.

Although the CSCT is a well-established standard, it is both qual-tative and subjective. We improved the interpretation of CSCT bynalyzing strips in a mathematical color space, specifically L*a*b*olor space (the most complete, perceptually linear color model)55]. We adapted the dimensionless L* axis of this space, whichescribes the “lightness” of an image, to measure the corrosion

f copper strips. Lightly tarnished strips generally have high L*alues (180–210), while severely tarnished strips generally haveow L* values (120–150). While the usual CSCT was designed for0 mL fluid samples and large copper strips (75 mm × 12.5 mm, upo 3.0 mm thick), we used very small, circular copper coupons that

A 1217 (2010) 2703–2715

fit in the bottom of GC autosampler vials. Moreover, the symmetriccircular geometry facilitates the analysis of the images with L*a*b*color space, and the small size of the coupons can actually facilitatecorrosion testing [56].

We applied the ADC approach to several crude oils, and sandcrude [57,58]. In a more exotic application, we measured a “crudeoil” made from swine manure [16]. To make the oil, swine manure,suspended in water, is pressurized in a reactor with CO and heatedto approximately 300 ◦C. The overall yield of oil from the reactor isapproximately 11% (mass/mass). In Fig. 12, we present a distillationcurve, along with the CSCT results. Insets show FTIR spectra of anearly and late fraction, and GC-MS of a mid-fraction. The relativelyhigh water content of this oil causes the distillation temperaturesto start at a low value and jump when the organics begin to distill.The high water content early in the distillation is reflected in theFTIR data, as is the high hydrocarbon content that develops later.The CSCT shows the fluid to somewhat corrosive through the dis-tillation curve. The L* values (not listed here) correlated with theCSCT ratings.

Analyses by GC-MS showed that the swine manure crude isa very complex mixture: even when investigating only the mainpeaks (with an abundance above 1%), 83 different organic com-pounds were identified. The main peaks from the low boiling regiondistillate samples were identified as nitrogenous heterocycles: sub-stituted pyrazines and pyrroles. Also identified in the first dropwere thiophenes. The sulfur in the thiophenes was also quanti-tated by GC-SCD. The high boiling fractions were dominated bylong-chain hydrocarbons; fluids from octane to octadecane wereidentified. In addition to these hydrocarbons, an interesting com-ponent identified on the basis of its mass spectrum was coprostane.Coprostane is the parent hydrocarbon of coprostanol (also calledcoprosterol, CAS No. 360-68-9), which is a main sterol found inswine fecal matter. Its presence indicates that the thermal conver-sion conditions of swine manure to crude oil were not sufficient tothermally crack this polycyclic compound.

Unlike our experiences with finished fuels or other crudes, alarge fraction of particulate char remained after distillation. TheADC allows recovery and analysis of this material. A powder X-raydiffraction pattern was inconclusive. Consequently, the char wasanalyzed with instrumental neutron activation analysis and coldneutron prompt gamma activation analysis. These complementaryneutron activation analysis techniques detected the presence of:Fe, Zn, Ag, Co, Cr, La, Sc, W, and very small amounts of Au and Hf.Metals such as Fe have been found elsewhere in swine manure andlagoon sludge.

3. Thermodynamic modeling

Thus far in the discussion, we have focused on the analyticalapplications of the ADC. An important contribution to the conceptof petroleomics (as advanced by Marshall and Rogers), however, isthe ability to use the information to advance the applied theory ofcomplex fluids so as to describe and predict the physical propertiesof the mixture and its components [59]. In their landmark review,Marshall and Rogers defined the term as the relationship betweenthe chemical composition of a fossil fuel and its properties and reac-tivity. Establishing this relationship has been more difficult than thedetailed analysis of the components in fossil fuels. Since the ADCproduces thermodynamically consistent temperatures along withthe relevant composition picture, it is ideally suited for the develop-

ment of such complex fluid theory. The basic idea in our approachis to represent the molar Helmholtz energy, a, of a mixture as a sumof an ideal contribution, aidsol, and an excess contribution aexcess:

a = aidsol + aexcess, (1)

T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715 2713

F by thee insets.

a

a

wancttF[�

ı

�

�

T

wsa

ig. 12. The distillation curve of the crude oil made from swine manure is shownarly and a late fraction, and GC-MS results for a middle fraction, are shown in the

idsol =m∑

j=1

xi[a0i (�, T) + ar

i (ı, �) + RT ln xi], (2)

excess = RT

m−1∑i=1

m∑j=i+1

xixjFij

∑k

Nkıdk �tk exp(−ılk ), (3)

here � and T are the mixture molar density and temperature, ınd � are the reduced mixture density and temperature, m is theumber of components, a0

iis the ideal gas Helmholtz energy of

omponent i, ari

is the residual Helmholtz energy of component i,he xi are the mole fractions of the constituents of the mixture, dk,k, lk and Nk are coefficients found from fitting experimental data,ij is an interaction parameter, and R is the universal gas constant60]. Mixing rules are used to determine the reducing parametersred and Tred for the mixture that are defined as

= �

�red, (4)

= Tred

T, (5)

red =

⎡⎣ m∑

i=1

xi

�ci

+m−1∑i=1

m∑j=i+1

xixj�ij

⎤⎦

−1

, (6)

=m∑

x Tc +m−1∑ m∑

x x ς , (7)

red

i=1

i i

i=1 j=i+1

i j ij

here �ij and ςij are binary interaction parameters that define thehapes of the reducing temperature and density curves, and Trednd �red are the reduced temperature and density.

diamonds, along with the CSCT coupons for each fraction. The FTIR spectra for an

The model has three binary interaction parameters for eachcomponent pair, �ij, ςij and Fij that can be determined by fittingexperimental data. If the constituent fluids are chemically sim-ilar, the excess contribution can be set to zero (i.e., Fij = 0), andthe �ij interaction parameter to zero, resulting in a simpler modelwith only one binary interaction parameter, ςij. Previous studieson refrigerant mixtures have shown that ςij is the most importantbinary parameter. This parameter can be found by fitting binarymixture data, or when data are unavailable, the following predictivescheme is used:

ςij = Tc2

Tc1

(40.4 − 25.03 × 2s) (8)

where

s =(

Tc1

Tc2

�c2

�c1

ω2

ω1

), (9)

where the fluid with the smaller dipole moment is designated asfluid “1”, and ω is the acentric factor [61].

The model for calculating the transport properties of a mixtureis an extended corresponding states method [62]. In this approach,the viscosity or thermal conductivity of a mixture is calculatedin a two-step procedure. First, mixing and combining rules areused to represent the mixture in terms of a hypothetical purefluid, then the properties of the hypothetical pure fluid are deter-mined by mapping onto a reference fluid through the use of “shapefactors”; details are given elsewhere. For both refrigerant mix-

tures and mixtures of natural gas components the viscosity andthermal conductivity are typically represented to within 5–10%.The two models discussed briefly above, the Helmholtz energymixing model for thermodynamic properties and the extended cor-responding states model for viscosity and thermal conductivity, are

2714 T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715

F n of sa thoutt te the

ics

fpumtt

wmfstdrtdddiu

dtpeclitchndtFpd

ft(

from the liquid phase is also delayed. The difference in the 10 and30% mixtures is dramatic, with the vaporization of n-nonane beingdelayed much later into the distillation than the 10% mixture. Thisresults in the distillation curves approaching, but never merging,even though the dimethyl carbonate itself has vaporized.

Table 1Constituents and mole fractions used in the simple surrogate models developed tosimulate the behavior of the distillation curves of diesel fuel with dimethyl carbonate(DMC).

Compound Mole fractioncomposition ofthe 10% (v/v)DMC mixture

Mole fractioncomposition of the20% (v/v) DMCmixture

Mole fractioncomposition ofthe 30% (v/v)DMC mixture

n-Nonane 0.0200 0.0160 0.0140n-Decane 0.0400 0.0310 0.0280n-Undecane 0.1000 0.0790 0.0700n-Dodecane 0.4046 0.3178 0.2848

ig. 13. A plot showing the distillation curves modeled with the Helmholtz equatioviation turbine fuel S-8, made from natural gas, and shows the model with and wiurbine fuel in which the Helmholtz equation of state is used predictively to genera

mplemented in NIST’s REFPROP computer program. This programontains highly accurate equations of state for pure fluids, includingome adopted as international standards [60].

We can use the theoretical formalism presented above in dif-erent ways. First, we can correlate experimental property data,roducing a model to represent the data within experimentalncertainty. Second, we can use the model predictively to esti-ate property values, based on limited experimental data. With

he ADC as a primary experimental input, we have used both ofhese approaches.

Returning to the synthetic aviation fuel S-8 discussed earlier,e can represent the composition of the fluid with a surrogateixture with components representing families of compounds

ound in S-8. Then, correlating measured density, heat capacity,ound speed, viscosity and thermal conductivity, it is possibleo model the properties of the mixture. Without the advancedistillation curve as an input, however, the ability of the model toepresent volatility is severely flawed [17]. We show in Fig. 13ahe experimental measurements, and calculated distillation curveseveloped with the Helmholtz model with and without the ADCata as an input. Including the distillation curve in the modelevelopment allows correlation of the volatility to within exper-

mental uncertainty, while failing to do so results in a physicallynrealistic representation.

We can also use the formalism presented above in a pre-ictive fashion, whereby we use a chemical analysis along withhe advanced distillation curve to predict the remaining physicalroperty information (density, sound speed and viscosity). As anxample, we study another synthetic substitute for JP-8, a blendedoal-derived fluid (CDF) made from a significant fraction of coaliquids and light cycle oil, a by-product of catalytic cracking unitsn petroleum refining [18,31]. The resulting mixture was treatedo increase the number of carbon–hydrogen bonds by hydropro-essing at high temperature and pressure. The fluid is intended forigh chemical stability up to 480 ◦C (900 ◦F, hence the alternativeame of the prototype, JP-900). The chemical analysis allowed theevelopment of a five-component surrogate, and the resulting mix-ure model calculated the distillation curve very well as shown inig. 13b. Moreover, the mixture model can represent other physicalroperties to within experimental uncertainty, realizing that such

ata are very limited.

In the discussion earlier on the oxygenating additives for dieseluel, we noted that in some cases, the additive has an effect onhe volatility even after the additive has been completely removedby distillation) from the mixture. This was illustrated with diethyl

tate as compared with experimental data. The plot shown in (a) is for the syntheticthe incorporation of the ADC data. The plot shown in (b) is for a coal-derived liquiddistillation curve.

carbonate, but we have observed the same effect with many morevolatile additives that vaporize early in the distillation. This occursbecause the energy being applied to the solution during the dis-tillation is being used to vaporize the additive, and the lightercomponents of the fluid undergo delayed vaporization. We can usethe thermodynamic models to demonstrate this, and predict thevaporization of the relevant species during the course of distilla-tion. To do this we construct a very simplified surrogate mixturefor diesel fuel and dimethyl carbonate (DMC), as listed in Table 1[32]. We choose DMC for this illustration, since the effect is dra-matic and easily demonstrated on a plot. We can calculate thedistillation curves for the three mixtures, and indeed the curveswill not completely merge, even after the DMC has vaporized.Fig. 14a and b shows further calculations from our surrogate model,specifically tracking how the composition of the liquid and vaporphases change as the distillation proceeds. The compositions ofDMC, n-nonane (the lightest component in the surrogate diesel)and n-hexadecane (the heaviest component in the surrogate diesel)are shown. Fig. 14a shows that the concentration of n-nonane isaffected significantly by the additive, while the n-hexadecane isaffected to a much lesser extent. This behavior is shown for two ini-tial concentrations of dimethyl carbonate (30% (v/v), and 10% (v/v))as described in Table 1. In Fig. 14a, the peak in the vapor phase con-centration of nonane is delayed, so that the removal of n-nonane

n-Tridecane 0.1000 0.0790 0.0700n-Tetradecane 0.0400 0.0310 0.0280n-Pentadecane 0.0400 0.0310 0.0280n-Hexadecane 0.0200 0.0160 0.0140Dimethyl carbonate 0.2354 0.3992 0.5368

T.J. Bruno et al. / J. Chromatogr. A 1217 (2010) 2703–2715 2715

F for thes ole fra

4

ctbmavbp

R

[

[[[[[[[[

[[[[

[[[[[

[

[[[[[[[[[[[[[[[[[[[

[

[

[[

[[[[

[[[[59] A.G. Marshall, R.P. Rodgers, Acc. Chem. Res. 37 (2004) 53.

ig. 14. (a) Calculated distillation curves (expressed as mole fractions of vapor, yi)urrogate summarized in Table 1. (b) Calculated distillation curves (expressed as m

. Conclusion

In this review, we have discussed the salient features of theomposition-explicit or advanced distillation curve approach forhe measurement of complex, multi-component fluids. The methodriges the gap between a chemical analysis protocol and a ther-ophysical property measurement in a relational manner. Thus,

nalytical information can be used to enhance a measure of fluidolatility, and vice versa. The result is a powerful method that cane used to characterize fluids and in the development of thermo-hysical property models.

eferences

[1] Separation and Purification; Critical Needs and Opportunities, Committee onSeparation Science and Technology, Board of Chemical Sciences and Technol-ogy, Commission on Physical Science, Mathematics and Resources, NationalResearch Council, 1987.

[2] J.R. Conder, C.L. Young, Physicochemical Measurements by Gas Chromatogra-phy, John Wiley and Sons, Chichester, 1979.

[3] H.Z. Kister, Distillation Operation, McGraw-Hill, New York, 1988.[4] H.Z. Kister, Distillation Design, McGraw-Hill, New York, 1991.[5] C.P. Angelos, S.V. Bhagwat, M.A. Matthews, Fluid Phase Equilib. 72 (1992) 182.[6] M.S. Peters, K.D. Timmerhaus, R.E. West, Plant Design and Economics for Chem-

ical Engineers, McGraw-Hill, New York, 2002.[7] M.R. Riazi, Characterization and Properties of Petroleum Fractions, ASTM

MNL50, American Society for Testing and Materials, West Conshohoken, PA,2005.

[8] M.R. Riazi, H.A. Al-Adwani, A. Bishara, J. Petrol. Sci. Eng. 42 (2004) 195.[9] D.A. Stark, Notes on Petroleum Processing, University of Notre Dame, 1998,

http://www.nd.edu/∼msen/Teaching/DirStudies/Petroleum.10] ASTM Standard D 86-04b, Standard Test Method for Distillation of Petroleum

Products at Atmospheric Pressure, ASTM International, West Conshohocken,PA, 2004.

11] T.J. Bruno, Ind. Eng. Chem. Res. 45 (2006) 4371.12] T.J. Bruno, B.L. Smith, Ind. Eng. Chem. Res. 45 (2006) 4381.13] T.J. Bruno, Sep. Sci. Technol. 41 (2006) 309.14] B.L. Smith, T.J. Bruno, Int. J. Thermophys. 27 (2006) 1419.15] T.J. Bruno, A. Wolk, A. Naydich, Energy Fuels 23 (2009) 3277.16] L.S. Ott, B.L. Smith, T.J. Bruno, Fuel 87 (2008) 3379.17] M.L. Huber, B.L. Smith, L.S. Ott, T.J. Bruno, Energy Fuels 22 (2008) 1104.

18] M.L. Huber, E.W. Lemmon, V. Diky, B.L. Smith, T.J. Bruno, Energy Fuels 22 (2008)

3249.19] M.L. Huber, E. Lemmon, L.S. Ott, T.J. Bruno, Energy Fuels 23 (2009) 3083.20] T.J. Bruno, Anal. Chem. 58 (1986) 1595.21] T.J. Bruno, J. Chem. Educ. 64 (1987) 987.22] T.J. Bruno, American Laboratory, Shelton, CT, United States, 1993, p. 16.

[[

[

mixture of 30% (v/v) dimethyl carbonate and 10% (v/v) dimethyl carbonate in thections of liquid, xi) for the surrogate.

23] A. Crafts, Berl. Berichte. 20 (1897) 709.24] L.S. Ott, B.L. Smith, T.J. Bruno, J. Chem. Thermodynam. 40 (2008) 1352.25] S. Young, Proc. Chem. Soc. 81 (1902) 777.26] S. Young, Fractional Distillation, Macmillan and Co., Ltd., London, 1903.27] S. Young, Distillation Principles and Processes, Macmillan and Co., Ltd., London,

1922.28] Purchase Description, RP-1, Federal Business Opportunities, http://fs2.eps.gov/

EPSData/DLA/Synopses/12671/SP0600-03-R-0322/rp-1sol.pdf, 2003.29] B.L. Smith, T.J. Bruno, Ind. Eng. Chem. Res. 46 (2007) 310.30] B.L. Smith, T.J. Bruno, J. Propul. Power 24 (2008) 619.31] B.L. Smith, T.J. Bruno, Energy Fuels 21 (2007) 2853.32] T.J. Bruno, A. Wolk, A. Naydich, M.L. Huber, Energy Fuels 23 (2009) 3989.33] L.S. Ott, B.L. Smith, T.J. Bruno, Energy Fuels 22 (2008) 2518.34] L.S. Ott, A. Hadler, T.J. Bruno, Ind. Eng. Chem. Res. 47 (2008) 9225.35] B.L. Smith, L.S. Ott, T.J. Bruno, Environ. Sci. Technol. 42 (2008) 7682.36] T.M. Lovestead, T.J. Bruno, Energy Fuels 23 (2009) 3637.37] B.L. Smith, T.J. Bruno, Ind. Eng. Chem. Res. 46 (2007) 297.38] T.J. Bruno, B.L. Smith, Energy Fuels 20 (2006) 2109.39] T.M. Lovestead, T.J. Bruno, Energy Fuels 23 (2009) 2176.40] T.J. Bruno, A. Wolk, A. Naydich, Energy Fuels 23 (2009) 2295.41] A.B. Hadler, L.S. Ott, T.J. Bruno, Fluid Phase Equilib. 281 (2009) 49.42] L.S. Ott, T.J. Bruno, Energy Fuels 22 (2008) 2861.43] B.L. Smith, L.S. Ott, T.J. Bruno, Ind. Eng. Chem. Res. 47 (2008) 5832.44] T.J. Bruno, A. Wolk, A. Naydich, Energy Fuels 23 (2009) 1015.45] T.J. Bruno, Process Control and Quality, 1992, p. 195.46] J.A. Widegren, T.J. Bruno, Energy Fuels, submitted for publication.47] J.A. Widegren, T.J. Bruno, Proceedings of 4th Liquid Propulsion Subcommittee,

JANNAF, December, 2008.48] ASTM D 130, Standard Test Method for Corrosiveness to Copper from Petroleum

Products by Copper Strip Test, ASTM International, West Conshohocken, PA,2004.

49] ASTM D 1838-07, Standard Test Method for Copper Strip Corrosion by LiquefiedPetroleum (LP) Gases, ASTM International, West Conshohocken, PA, 2007.

50] W.C. Andersen, A.I. Abdulagatov, T.J. Bruno, Energy Fuels 17 (2003) 120.51] W.C. Andersen, T.J. Bruno, Abstracts of Papers 225th ACS National Meeting, New

Orleans, LA, United States, 23–27 March, ETR, 2003.52] W.C. Andersen, T.J. Bruno, Ind. Eng. Chem. Res. 42 (2003) 971.53] W.C. Andersen, T.J. Bruno, Ind. Eng. Chem. Res. 42 (2003) 963.54] P.D.N. Svoronos, T.J. Bruno, Ind. Eng. Chem. Res. 41 (2002) 5321.55] W.C. Andersen, G.C. Straty, T.J. Bruno, Proceedings of the GTI natural Gas Tech-

nologies Conference and Exhibition, Phoenix, AZ, 2004.56] L.S. Ott, T.J. Bruno, J. Sulfur Chem. 28 (2007) 493.57] L.S. Ott, T.J. Bruno, Energy Fuels 21 (2007) 2778.58] L.S. Ott, B.L. Smith, T.J. Bruno, Fuel 87 (2008) 3055.

60] E.W. Lemmon, R.T. Jacobsen, J. Phys. Chem. Ref. Data 33 (2004) 593.61] E.W.M. Lemmon, M.O. McLinden, Proceedings of the Thermophysical Properties

and Transfer Processes of New Refrigerants Conference, International Instituteof Refrigeration, Paderborn, Germany, 2001, p. 23.

62] M.L. Huber, A. Lasecke, R.A. Perkins, Ind. Eng. Chem. Res. 42 (2003) 3163.